[Beowulf] The Walmart Compute Node?
Robert G. Brown
rgb at phy.duke.edu
Fri Nov 9 05:00:48 PST 2007
On Thu, 8 Nov 2007, Jim Lux wrote:
> At 01:03 PM 11/8/2007, andrew holway wrote:
>> Im still not convinced, bang for buck your going to get more
>> clustering this junk than buying commodity hardware. Benchmarks at the
> No question there..
> However, if you want a low capital investment toy cluster to learn on, the
> $200 node is quite attractive.
> For instance, say you had a class of 10 people wanting to learn about cluster
> design and admin. You could go and get a big cluster and install the
> necessary management tools to let them share the cluster, or use an already
> existing cluster. OTOH, you could also buy 10 $2K clusters, and let them
> have at it on their very own. They'll learn all the things about cluster
> admining, rebuilding images, etc.
Now THIS I completely agree with. Nor is it a crazy situation -- I help
(indeed, am actively helping) high school students with cluster designs
for their high school project(s), and while they usually end up using
recycled old heterogeneous junk at least at first, there would be a real
benefit to putting together a small pile of brand new systems.
However, multicores are starting to redefine the optimum even here.
When one can get a dual core system for $400 (with a higher clock and a
64 bit CPU) then getting two BOXES at $200 each is still not horribly
attractive compared to getting one BOX at $400 that is probably close to
twice the aggregate flops. So instead of getting 8 boxes for $1600 +
$250-400 for a monitor, KVM, cables, switch, and shelving you get only 4
boxes (but have 10x the network bandwidth per box and actually win
tremendously on scalability) for your eight processors. You are
"forced" to learn about scalability of tasks across an inhomogeneous IPC
landscape quite early, but that isn't necessarily a bad thing...
> For this kind of thing, raw performance isn't important, number of nodes is,
> because there's a qualitative difference between running a standalone machine
> and running a group of N>3 machines that have to talk. Somewhere around 5-10
> computers, brute force techniques to install software like sneakernet start
> to really break down.
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit
Robert G. Brown
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Lulu Bookstore: http://stores.lulu.com/store.php?fAcctID=877977
More information about the Beowulf